docker-compose.yml

# ⚠️  TimescaleDB-HA migration (ADR-007):
#   This compose file was migrated to timescale/timescaledb-ha:pg17.
#   If upgrading from a postgres:17-alpine deployment, you MUST wipe the volume:
#       docker compose down
#       docker volume rm teslasync_postgres_data
#       docker compose up -d
#   Existing data is replaced; staging losses accepted per ADR-009.

services:
  teslasync-api:
    build:
      context: .
      dockerfile: Dockerfile
      args:
        VERSION: ${VERSION:-dev}
    container_name: teslasync-api
    restart: unless-stopped
    ports:
      - "${TESLASYNC_PORT:-8080}:8080"
    environment:
      - TESLASYNC_PORT=8080
      - DATABASE_HOST=postgres
      - DATABASE_PORT=5432
      - DATABASE_USER=${POSTGRES_USER:-teslasync}
      - DATABASE_PASSWORD=${POSTGRES_PASSWORD:-teslasync}
      - DATABASE_NAME=${POSTGRES_DB:-teslasync}
      - DATABASE_SSL_MODE=disable
      - DATABASE_CONNECT_TIMEOUT=5
      - DATABASE_STATEMENT_TIMEOUT=30000
      - DATABASE_HEALTH_CHECK_PERIOD=5s
      - MQTT_HOST=mosquitto
      - MQTT_PORT=1883
      - REDIS_HOST=redis
      - REDIS_PORT=6379
      - REDIS_ENABLED=true
      - TESLA_CLIENT_ID=${TESLA_CLIENT_ID}
      - TESLA_CLIENT_SECRET=${TESLA_CLIENT_SECRET}
      - TESLA_REDIRECT_URI=${TESLA_REDIRECT_URI:-http://localhost:8080/api/v1/auth/callback}
      - TESLA_API_BASE_URL=${TESLA_API_BASE_URL:-https://fleet-api.prd.na.vn.cloud.tesla.com}
      - WORKER_POLL_INTERVAL=${WORKER_POLL_INTERVAL:-30s}
      - LOG_LEVEL=${LOG_LEVEL:-error}
      - FLEET_TELEMETRY_ENABLED=${FLEET_TELEMETRY_ENABLED:-false}
      - FLEET_TELEMETRY_HOST=${FLEET_TELEMETRY_HOST:-}
      - FLEET_TELEMETRY_PORT=${FLEET_TELEMETRY_PORT:-4443}
      - FLEET_TELEMETRY_TOPIC_BASE=${FLEET_TELEMETRY_TOPIC_BASE:-telemetry}
      - FLEET_TELEMETRY_BATCH_MS=${FLEET_TELEMETRY_BATCH_MS:-100}
      - FLEET_TELEMETRY_STALE_TIMEOUT=${FLEET_TELEMETRY_STALE_TIMEOUT:-15m}
      - FLEET_TELEMETRY_FALLBACK_POLL_INTERVAL=${FLEET_TELEMETRY_FALLBACK_POLL_INTERVAL:-5m}
      - FLEET_TELEMETRY_SNAPSHOT_WRITE_INTERVAL=${FLEET_TELEMETRY_SNAPSHOT_WRITE_INTERVAL:-1s}
      - LIVE_SIGNAL_STORE_MODE=${LIVE_SIGNAL_STORE_MODE:-hybrid}
      - OTEL_ENABLED=${OTEL_ENABLED:-false}
      - OTEL_ENDPOINT=${OTEL_ENDPOINT:-otel-collector:4317}
      - OTEL_EXPORTER_OTLP_ENDPOINT=${OTEL_EXPORTER_OTLP_ENDPOINT:-http://otel-collector:4317}
      - OTEL_TRACES_SAMPLER_ARG=${OTEL_TRACES_SAMPLER_ARG:-1.0}
      - OTEL_SERVICE_NAME=${OTEL_SERVICE_NAME:-teslasync-api}
      - OTEL_INSECURE=${OTEL_INSECURE:-true}
      # Pyroscope continuous profiling — opt-in via PYROSCOPE_ENABLED.
      # When enabled the API uploads CPU/heap/goroutine/mutex/block deltas
      # to the Pyroscope server every PYROSCOPE_UPLOAD_RATE (default 15s).
      # Phase-49 / p49-profiling.
      - PYROSCOPE_ENABLED=${PYROSCOPE_ENABLED:-false}
      - PYROSCOPE_SERVER_ADDRESS=${PYROSCOPE_SERVER_ADDRESS:-http://pyroscope:4040}
      - PYROSCOPE_UPLOAD_RATE=${PYROSCOPE_UPLOAD_RATE:-15s}
      # Phase-46 / SOTA observability batch.
      # SLO board — empty PROMETHEUS_BASE_URL serves catalog
      # metadata only (no live tier evaluation).
      - SLO_CATALOG_PATH=${SLO_CATALOG_PATH:-slo/catalog.yaml}
      - PROMETHEUS_BASE_URL=${PROMETHEUS_BASE_URL:-}
      # Data quality scoring over signal_log.
      - DATA_QUALITY_ENABLED=${DATA_QUALITY_ENABLED:-true}
      - DATA_QUALITY_WINDOW_MINS=${DATA_QUALITY_WINDOW_MINS:-60}
      # Synthetic monitoring (outside-in probes). Opt-in.
      - SYNTHETIC_ENABLED=${SYNTHETIC_ENABLED:-false}
      - SYNTHETIC_INTERVAL_SECONDS=${SYNTHETIC_INTERVAL_SECONDS:-60}
      - SYNTHETIC_TIMEOUT_SECONDS=${SYNTHETIC_TIMEOUT_SECONDS:-30}
      - SYNTHETIC_PROBE_URLS=${SYNTHETIC_PROBE_URLS:-}
      # Phase-44 / observability-batch / Prompt F4 — DLQ Inspector.
      # Replay is opt-in because re-publishing a payload that already
      # triggered a production exception is consequential. Replay is
      # also gated by sudo-token AND audited on every code path.
      - DLQ_REPLAY_ENABLED=${DLQ_REPLAY_ENABLED:-false}
      - DLQ_RING_CAPACITY=${DLQ_RING_CAPACITY:-200}
      - GOOGLE_MAPS_API_KEY=${GOOGLE_MAPS_API_KEY:-}
      - AZURE_MAPS_API_KEY=${AZURE_MAPS_API_KEY:-}
      # Inbound api_call_logs middleware (Phase 38-10).
      # Body capture defaults to OFF; enable only for diagnostic windows.
      - API_LOGS_INBOUND_ENABLED=${API_LOGS_INBOUND_ENABLED:-true}
      - API_LOG_CAPTURE_BODIES=${API_LOG_CAPTURE_BODIES:-false}
      - API_LOG_QUEUE_CAPACITY=${API_LOG_QUEUE_CAPACITY:-4096}
      - API_LOG_BATCH_SIZE=${API_LOG_BATCH_SIZE:-100}
      - API_LOG_FLUSH_INTERVAL=${API_LOG_FLUSH_INTERVAL:-1s}
      # Audit log retention (Phase-40 / Prompt 49 — Recent Activity Discoverability).
      # AUDIT_RETENTION_DAYS: hard delete entries older than N days (0 = never).
      # AUDIT_IP_RETENTION_DAYS: redact ip + user_agent on entries older than N days
      # so per-user activity stays usable without retaining network PII forever.
      - AUDIT_RETENTION_DAYS=${AUDIT_RETENTION_DAYS:-365}
      - AUDIT_IP_RETENTION_DAYS=${AUDIT_IP_RETENTION_DAYS:-30}
      # Web Push (VAPID) — Phase 40 / Prompt 52.
      # Generate keys once with: go run ./cmd/teslasync vapid-keygen
      # When unset, the API logs a Warn at startup and the push channel
      # is disabled (no behaviour change for existing installs).
      - TESLASYNC_VAPID_PUBLIC_KEY=${TESLASYNC_VAPID_PUBLIC_KEY:-}
      - TESLASYNC_VAPID_PRIVATE_KEY=${TESLASYNC_VAPID_PRIVATE_KEY:-}
      - TESLASYNC_VAPID_SUBJECT=${TESLASYNC_VAPID_SUBJECT:-}
      # Service-mode banner override — Phase 46 / Prompt 04.
      # When TESLASYNC_SYSTEM_MODE is set ("ok", "degraded",
      # "maintenance"), it overrides the system_state DB row so an
      # operator can force-set or force-clear the top-of-app banner
      # during deploys/rollbacks without touching the database. Leave
      # empty to defer to the DB row + admin POST API.
      - TESLASYNC_SYSTEM_MODE=${TESLASYNC_SYSTEM_MODE:-}
      - TESLASYNC_SYSTEM_MAINTENANCE_MESSAGE=${TESLASYNC_SYSTEM_MAINTENANCE_MESSAGE:-}
      - TESLASYNC_SYSTEM_MAINTENANCE_UNTIL=${TESLASYNC_SYSTEM_MAINTENANCE_UNTIL:-}
      # Optional GitHub Issues bridge for the in-app feedback widget
      # (Phase 46 / Prompt 08). When both REPO ("owner/name") and TOKEN
      # (PAT with Issues:write) are set, the admin feedback queue
      # exposes a "Forward to GitHub" action that mirrors the user
      # report into a real GitHub issue. Leave empty to disable — the
      # SPA hides the action when the server reports the bridge as
      # unconfigured.
      - TESLASYNC_GITHUB_REPO=${TESLASYNC_GITHUB_REPO:-}
      - TESLASYNC_GITHUB_TOKEN=${TESLASYNC_GITHUB_TOKEN:-}
      # Vehicle photo upload root — Phase 46 / Prompt 54.
      # User-uploaded vehicle hero photos are written to this dir
      # at runtime (3 sizes per upload). Mount a volume here in
      # production so photos survive container restart.
      - TESLASYNC_VEHICLE_PHOTO_DIR=${TESLASYNC_VEHICLE_PHOTO_DIR:-/var/lib/teslasync/photos}
      # Auth-mode contract — Phase 46 / Prompt 57.
      # Free-text hint surfaced verbatim by /system/auth-mode and
      # rendered by the SPA's RequiresAuth empty state. Set to the
      # name of your upstream IdP (Authentik, Authelia, oauth2-proxy,
      # Keycloak, …) for nicer copy. TeslaSync NEVER speaks to the
      # IdP's admin API and NEVER routes off this value — it is
      # presentation-only. Default empty: SPA falls back to a
      # generic "your authentication provider" string.
      - TESLASYNC_AUTH_PROVIDER_HINT=${TESLASYNC_AUTH_PROVIDER_HINT:-}
      # Cookie / GDPR consent banner — Phase 46 / Prompt 70.
      # Default OFF: self-hosted single-user installs do not need a
      # consent banner. Flip to "true" on multi-user fleet
      # deployments or any public-facing instance with optional
      # client-side reporting enabled (web vitals, error reporter)
      # to comply with ePrivacy / GDPR. When true, the SPA renders
      # a non-blocking banner the first visit and gates optional
      # POSTs on the user's stored consent.
      - TESLASYNC_REQUIRE_COOKIE_CONSENT=${TESLASYNC_REQUIRE_COOKIE_CONSENT:-false}
    volumes:
      - vehicle_photos:/var/lib/teslasync/photos
    depends_on:
      postgres:
        condition: service_healthy
      mosquitto:
        condition: service_started
      redis:
        condition: service_healthy
    healthcheck:
      test: ["CMD", "/usr/local/bin/teslasync", "healthcheck"]
      interval: 30s
      timeout: 5s
      retries: 3
      start_period: 10s
    deploy:
      resources:
        limits:
          memory: 512M
        reservations:
          memory: 128M

  web:
    build:
      context: .
      dockerfile: Dockerfile.web
      # Frontend RUM is build-time inlined: Vite reads VITE_* at build to bake
      # them into the bundle. Forwarding them as build args means a single
      # `docker compose build web` picks up changes to the collector endpoint
      # without re-engineering the Dockerfile. The SPA gracefully no-ops when
      # VITE_OTLP_HTTP_ENDPOINT is unset (see web/src/observability/rum.ts).
      args:
        VITE_OTLP_HTTP_ENDPOINT: ${VITE_OTLP_HTTP_ENDPOINT:-}
        VITE_OTEL_SERVICE_NAME: ${VITE_OTEL_SERVICE_NAME:-teslasync-web}
        VITE_OTEL_DEPLOY_ENV: ${VITE_OTEL_DEPLOY_ENV:-dev-compose}
        VITE_APP_VERSION: ${VERSION:-dev}
    container_name: teslasync-web
    restart: unless-stopped
    ports:
      - "${WEB_PORT:-3000}:80"
    depends_on:
      teslasync-api:
        condition: service_healthy
    deploy:
      resources:
        limits:
          memory: 128M
        reservations:
          memory: 32M

  notification-worker:
    build:
      context: .
      dockerfile: Dockerfile.notification
      args:
        VERSION: ${VERSION:-dev}
    container_name: teslasync-notification-worker
    restart: unless-stopped
    environment:
      - DATABASE_HOST=postgres
      - DATABASE_PORT=5432
      - DATABASE_USER=${POSTGRES_USER:-teslasync}
      - DATABASE_PASSWORD=${POSTGRES_PASSWORD:-teslasync}
      - DATABASE_NAME=${POSTGRES_DB:-teslasync}
      - DATABASE_SSL_MODE=disable
      - MQTT_HOST=mosquitto
      - MQTT_PORT=1883
      - LOG_LEVEL=${LOG_LEVEL:-error}
      - HEALTH_PORT=8081
      - OTEL_ENABLED=${OTEL_ENABLED:-false}
      - OTEL_ENDPOINT=${OTEL_ENDPOINT:-otel-collector:4317}
      - OTEL_SERVICE_NAME=${OTEL_SERVICE_NAME:-teslasync-notification-worker}
      - OTEL_INSECURE=${OTEL_INSECURE:-true}
      - PYROSCOPE_ENABLED=${PYROSCOPE_ENABLED:-false}
      - PYROSCOPE_SERVER_ADDRESS=${PYROSCOPE_SERVER_ADDRESS:-http://pyroscope:4040}
      - PYROSCOPE_UPLOAD_RATE=${PYROSCOPE_UPLOAD_RATE:-15s}
      # Web Push (VAPID) — same env vars as the API server. Required so
      # the notification worker can deliver "webpush" Requests pulled
      # from the MQTT internal topic.
      - TESLASYNC_VAPID_PUBLIC_KEY=${TESLASYNC_VAPID_PUBLIC_KEY:-}
      - TESLASYNC_VAPID_PRIVATE_KEY=${TESLASYNC_VAPID_PRIVATE_KEY:-}
      - TESLASYNC_VAPID_SUBJECT=${TESLASYNC_VAPID_SUBJECT:-}
    depends_on:
      postgres:
        condition: service_healthy
      mosquitto:
        condition: service_started
    healthcheck:
      test: ["CMD", "/usr/local/bin/notification-worker", "healthcheck"]
      interval: 30s
      timeout: 5s
      retries: 3
      start_period: 10s
    deploy:
      resources:
        limits:
          memory: 128M
        reservations:
          memory: 32M

  postgres:
    # ⚠️  TimescaleDB-HA migration (ADR-007):
    #   Migrated from postgres:17-alpine to timescale/timescaledb-ha:pg17.
    #   If upgrading an existing deployment, you MUST wipe the volume:
    #       docker compose down
    #       docker volume rm teslasync_postgres_data
    #       docker compose up -d
    #   Existing data is replaced; staging losses accepted per ADR-009.
    image: timescale/timescaledb-ha:pg17
    container_name: teslasync-postgres
    restart: unless-stopped
    command: ["postgres", "-c", "max_connections=100", "-c", "shared_buffers=256MB", "-c", "shared_preload_libraries=timescaledb,pg_stat_statements,pg_textsearch", "-c", "pg_stat_statements.track=all"]
    ports:
      - "${POSTGRES_PORT:-5432}:5432"
    environment:
      - POSTGRES_USER=${POSTGRES_USER:-teslasync}
      - POSTGRES_PASSWORD=${POSTGRES_PASSWORD:-teslasync}
      - POSTGRES_DB=${POSTGRES_DB:-teslasync}
      - PGDATA=/home/postgres/pgdata/data
    volumes:
      - postgres_data:/home/postgres/pgdata
      - ./scripts/init-timescaledb.sql:/docker-entrypoint-initdb.d/00-init-timescaledb.sql:ro
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U ${POSTGRES_USER:-teslasync}"]
      interval: 10s
      timeout: 5s
      retries: 5
    deploy:
      resources:
        limits:
          memory: 1G
        reservations:
          memory: 256M

  grafana:
    image: grafana/grafana:10.4.0
    container_name: teslasync-grafana
    restart: unless-stopped
    ports:
      - "${GRAFANA_PORT:-3001}:3000"
    environment:
      - GF_SECURITY_ADMIN_USER=${GRAFANA_USER:-admin}
      - GF_SECURITY_ADMIN_PASSWORD=${GRAFANA_PASSWORD:-teslasync}
      - GF_INSTALL_PLUGINS=grafana-clock-panel
      - POSTGRES_HOST=postgres
      - POSTGRES_PORT=5432
      - POSTGRES_USER=${POSTGRES_USER:-teslasync}
      - POSTGRES_PASSWORD=${POSTGRES_PASSWORD:-teslasync}
      - POSTGRES_DB=${POSTGRES_DB:-teslasync}
      - PROMETHEUS_URL=http://prometheus:9090
    volumes:
      - grafana_data:/var/lib/grafana
      - ./grafana/provisioning:/etc/grafana/provisioning
      - ./grafana/dashboards:/var/lib/grafana/dashboards
    depends_on:
      postgres:
        condition: service_healthy
      prometheus:
        condition: service_started
    deploy:
      resources:
        limits:
          memory: 256M
        reservations:
          memory: 64M

  prometheus:
    image: prom/prometheus:v2.51.0
    container_name: teslasync-prometheus
    restart: unless-stopped
    ports:
      - "${PROMETHEUS_PORT:-9099}:9090"
    volumes:
      - ./prometheus/prometheus.yml:/etc/prometheus/prometheus.yml:ro
      # SLO recording + alerting rules are generated from slo/catalog.yaml by
      # `go run ./cmd/slogen generate {recording,alerting}` and committed under
      # helm/teslasync/files/prometheus/. Mounting the same artefacts here keeps
      # dev and K8s on identical rule definitions — drift would mean a SLO
      # tested in dev does not exist in prod.
      - ./helm/teslasync/files/prometheus/recording-rules.yaml:/etc/prometheus/rules/recording-rules.yaml:ro
      - ./helm/teslasync/files/prometheus/alerting-rules.yaml:/etc/prometheus/rules/alerting-rules.yaml:ro
      - prometheus_data:/prometheus
    command:
      - '--config.file=/etc/prometheus/prometheus.yml'
      - '--storage.tsdb.retention.time=15d'
      - '--web.enable-lifecycle'
      # exemplar-storage retains the trace_id/span_id labels attached to
      # histogram observations by internal/api/middleware.go +
      # internal/metrics/exemplar.go so Grafana can render the "View trace"
      # dot on RED latency panels. Without this flag the labels are accepted
      # at ingest but dropped at query time, hiding the link entirely.
      - '--enable-feature=exemplar-storage'
    depends_on:
      teslasync-api:
        condition: service_healthy
    deploy:
      resources:
        limits:
          memory: 256M
        reservations:
          memory: 64M

  mosquitto:
    image: eclipse-mosquitto:2
    container_name: teslasync-mosquitto
    restart: unless-stopped
    ports:
      - "${MQTT_PORT:-1883}:1883"
      - "9001:9001"
    volumes:
      - mosquitto_data:/mosquitto/data
      - ./mosquitto.conf:/mosquitto/config/mosquitto.conf
    healthcheck:
      test: ["CMD", "mosquitto_sub", "-t", "$$SYS/#", "-C", "1", "-i", "healthcheck", "-W", "3"]
      interval: 30s
      timeout: 5s
      retries: 3
    deploy:
      resources:
        limits:
          memory: 64M
        reservations:
          memory: 16M

  redis:
    image: redis:7-alpine
    container_name: teslasync-redis
    restart: unless-stopped
    ports:
      - "${REDIS_PORT:-6379}:6379"
    volumes:
      - redis_data:/data
    command: redis-server --appendonly yes --maxmemory 128mb --maxmemory-policy allkeys-lru
    healthcheck:
      test: ["CMD", "redis-cli", "ping"]
      interval: 10s
      timeout: 5s
      retries: 3
    deploy:
      resources:
        limits:
          memory: 192M
        reservations:
          memory: 32M

  export-worker:
    build:
      context: .
      dockerfile: Dockerfile.export-worker
      args:
        VERSION: ${VERSION:-dev}
    container_name: teslasync-export-worker
    restart: unless-stopped
    environment:
      - DATABASE_HOST=postgres
      - DATABASE_PORT=5432
      - DATABASE_USER=${POSTGRES_USER:-teslasync}
      - DATABASE_PASSWORD=${POSTGRES_PASSWORD:-teslasync}
      - DATABASE_NAME=${POSTGRES_DB:-teslasync}
      - DATABASE_SSL_MODE=disable
      - MQTT_HOST=mosquitto
      - MQTT_PORT=1883
      - LOG_LEVEL=${LOG_LEVEL:-error}
      - HEALTH_PORT=8082
      - OTEL_ENABLED=${OTEL_ENABLED:-false}
      - OTEL_ENDPOINT=${OTEL_ENDPOINT:-otel-collector:4317}
      - OTEL_SERVICE_NAME=${OTEL_SERVICE_NAME:-teslasync-export-worker}
      - OTEL_INSECURE=${OTEL_INSECURE:-true}
      - PYROSCOPE_ENABLED=${PYROSCOPE_ENABLED:-false}
      - PYROSCOPE_SERVER_ADDRESS=${PYROSCOPE_SERVER_ADDRESS:-http://pyroscope:4040}
      - PYROSCOPE_UPLOAD_RATE=${PYROSCOPE_UPLOAD_RATE:-15s}
    depends_on:
      postgres:
        condition: service_healthy
      mosquitto:
        condition: service_started
    healthcheck:
      test: ["CMD", "/usr/local/bin/export-worker", "healthcheck"]
      interval: 30s
      timeout: 5s
      retries: 3
      start_period: 10s
    deploy:
      resources:
        limits:
          memory: 512M
        reservations:
          memory: 64M

  automation-worker:
    build:
      context: .
      dockerfile: Dockerfile.automation
      args:
        VERSION: ${VERSION:-dev}
    container_name: teslasync-automation-worker
    restart: unless-stopped
    environment:
      - DATABASE_HOST=postgres
      - DATABASE_PORT=5432
      - DATABASE_USER=${POSTGRES_USER:-teslasync}
      - DATABASE_PASSWORD=${POSTGRES_PASSWORD:-teslasync}
      - DATABASE_NAME=${POSTGRES_DB:-teslasync}
      - DATABASE_SSL_MODE=disable
      - MQTT_HOST=mosquitto
      - MQTT_PORT=1883
      - TESLA_CLIENT_ID=${TESLA_CLIENT_ID}
      - TESLA_CLIENT_SECRET=${TESLA_CLIENT_SECRET}
      - TESLA_API_BASE_URL=${TESLA_API_BASE_URL:-https://fleet-api.prd.na.vn.cloud.tesla.com}
      - LOG_LEVEL=${LOG_LEVEL:-error}
      - HEALTH_PORT=8083
      - OTEL_ENABLED=${OTEL_ENABLED:-false}
      - OTEL_ENDPOINT=${OTEL_ENDPOINT:-otel-collector:4317}
      - OTEL_SERVICE_NAME=${OTEL_SERVICE_NAME:-teslasync-automation-worker}
      - OTEL_INSECURE=${OTEL_INSECURE:-true}
      - PYROSCOPE_ENABLED=${PYROSCOPE_ENABLED:-false}
      - PYROSCOPE_SERVER_ADDRESS=${PYROSCOPE_SERVER_ADDRESS:-http://pyroscope:4040}
      - PYROSCOPE_UPLOAD_RATE=${PYROSCOPE_UPLOAD_RATE:-15s}
    depends_on:
      postgres:
        condition: service_healthy
      mosquitto:
        condition: service_started
    healthcheck:
      test: ["CMD", "/usr/local/bin/automation-worker", "healthcheck"]
      interval: 30s
      timeout: 5s
      retries: 3
      start_period: 10s
    deploy:
      resources:
        limits:
          memory: 256M
        reservations:
          memory: 64M

  # Jaeger (optional — docker compose --profile tracing up)
  # Provides distributed tracing via OpenTelemetry (OTLP gRPC).
  # Jaeger UI: http://localhost:16686
  #
  # In the dev compose layout, services emit spans to the otel-collector below
  # rather than direct to jaeger, so the collector's batching + tail-sampling
  # (when enabled) takes effect even locally. The collector then forwards to
  # jaeger:4317 via the otlp/jaeger exporter. Jaeger no longer needs to bind
  # 4317 to the host — the host-facing OTLP ingest is owned by otel-collector.
  jaeger:
    image: jaegertracing/all-in-one:1.57
    container_name: teslasync-jaeger
    restart: unless-stopped
    profiles:
      - tracing
    ports:
      - "${JAEGER_UI_PORT:-16686}:16686"
    environment:
      - COLLECTOR_OTLP_ENABLED=true
    deploy:
      resources:
        limits:
          memory: 256M

  # OpenTelemetry Collector (optional — docker compose --profile tracing up)
  # Dev variant of helm/teslasync/files/otel-collector/config.yaml — 100%
  # sampling (no tail policy), CORS-enabled OTLP HTTP receiver for browser
  # RUM, and Jaeger as the trace backend instead of Tempo. See
  # observability/otel-collector/config.yaml for the full pipeline and the
  # rationale behind each dev-specific divergence.
  #
  # Ports:
  #   4317 - OTLP gRPC (Go services)
  #   4318 - OTLP HTTP (browser RUM, with CORS)
  #   9464 - Prometheus exposition of collector + forwarded OTLP metrics
  #   8888 - Collector's own internal telemetry metrics
  otel-collector:
    image: otel/opentelemetry-collector-contrib:0.103.1
    container_name: teslasync-otel-collector
    restart: unless-stopped
    profiles:
      - tracing
    command:
      - "--config=/etc/otelcol/config.yaml"
    volumes:
      - ./observability/otel-collector/config.yaml:/etc/otelcol/config.yaml:ro
    ports:
      - "${OTEL_COLLECTOR_GRPC_PORT:-4317}:4317"
      - "${OTEL_COLLECTOR_HTTP_PORT:-4318}:4318"
      - "${OTEL_COLLECTOR_METRICS_PORT:-9464}:9464"
    depends_on:
      jaeger:
        condition: service_started
    deploy:
      resources:
        limits:
          memory: 256M
        reservations:
          memory: 64M

  # Pyroscope continuous profiling server (optional — docker compose --profile profiling up)
  # Receives godeltaprof uploads from API + 3 workers. Surfaces CPU, heap,
  # goroutine, mutex, block profiles in Grafana via the Pyroscope datasource.
  # Phase-49 / p49-profiling. See docs/runbooks/phase-49-pyroscope.md.
  #
  # Ports:
  #   4040 - Pyroscope HTTP ingest + UI
  pyroscope:
    image: grafana/pyroscope:1.7.1
    container_name: teslasync-pyroscope
    restart: unless-stopped
    profiles:
      - profiling
    ports:
      - "${PYROSCOPE_PORT:-4040}:4040"
    command:
      - "server"
    volumes:
      - pyroscope_data:/var/lib/pyroscope
    deploy:
      resources:
        limits:
          memory: 512M
        reservations:
          memory: 128M

  # Fleet Telemetry Server (optional — docker compose --profile telemetry up)
  # Receives streaming data from Tesla vehicles via WebSocket and publishes
  # to MQTT (Mosquitto). TeslaSync's MQTT subscriber picks up the signals.
  # Requires TLS certificates and Tesla Developer account with Fleet Telemetry access.
  # See: https://github.com/teslamotors/fleet-telemetry
  fleet-telemetry:
    image: tesla/fleet-telemetry:latest
    container_name: teslasync-fleet-telemetry
    restart: unless-stopped
    profiles:
      - telemetry
    ports:
      - "${FLEET_TELEMETRY_PORT:-4443}:4443"
    volumes:
      - ./fleet-telemetry-config.json:/etc/fleet-telemetry/config.json:ro
      - ${FLEET_TELEMETRY_TLS_CERT:-./certs/server.crt}:/certs/server.crt:ro
      - ${FLEET_TELEMETRY_TLS_KEY:-./certs/server.key}:/certs/server.key:ro
    depends_on:
      mosquitto:
        condition: service_started
    deploy:
      resources:
        limits:
          memory: 256M
        reservations:
          memory: 64M

  # Vehicle Command Proxy (optional — enables signed vehicle commands)
  # Routes commands through Tesla's Vehicle Command SDK for end-to-end signing.
  # Required for 2021+ Model S/X and all Model 3/Y vehicles.
  # Setup: 1) Generate keypair via /dev-tools/generate-keypair
  #        2) Save private key to ./certs/private-key.pem
  #        3) Register public key with Tesla and pair with vehicle
  #        4) Set TESLA_COMMAND_PROXY_URL=https://vehicle-command-proxy:4443 in .env
  vehicle-command-proxy:
    image: tesla/vehicle-command:latest
    container_name: teslasync-vehicle-command-proxy
    restart: unless-stopped
    profiles:
      - commands
    environment:
      - TESLA_HTTP_PROXY_HOST=0.0.0.0
      - TESLA_HTTP_PROXY_PORT=4443
      - TESLA_HTTP_PROXY_TLS_CERT=/data/tls-cert.pem
      - TESLA_HTTP_PROXY_TLS_KEY=/data/tls-key.pem
      - TESLA_KEY_FILE=/data/private-key.pem
      - TESLA_VERBOSE=${TESLA_VERBOSE:-false}
    volumes:
      - ${COMMAND_PROXY_DATA_DIR:-./data/vehicle-command}:/data
    healthcheck:
      test: ["CMD", "wget", "--no-check-certificate", "-qO-", "https://localhost:4443/health"]
      interval: 30s
      timeout: 5s
      retries: 3
      start_period: 10s
    deploy:
      resources:
        limits:
          memory: 128M
        reservations:
          memory: 32M

  # Toxiproxy chaos / fault-injection harness (optional — docker compose --profile chaos up)
  # Phase-49 / p49-chaos. Programmable TCP proxy that injects latency,
  # drops, and bandwidth limits between TeslaSync and its backing
  # services. The cmd/chaos-runner binary drives it via the admin API
  # on :8474. See internal/chaos/scenarios.go for the default suite.
  #
  # NOT used in steady-state operation — operators bring this profile
  # up to exercise the recovery contracts of the FSM, signal pipeline,
  # and reconciliation loop. The runner removes every injected toxic
  # on completion; restart toxiproxy to clear any state if a run is
  # killed mid-execution.
  #
  # Ports:
  #   8474 - Toxiproxy HTTP admin API (host-bound for the runner)
  toxiproxy:
    image: ghcr.io/shopify/toxiproxy:2.9.0
    container_name: teslasync-toxiproxy
    restart: unless-stopped
    profiles:
      - chaos
    ports:
      - "${TOXIPROXY_PORT:-8474}:8474"
    command: ["-host=0.0.0.0", "-config=/etc/toxiproxy/config.json"]
    volumes:
      - ./observability/toxiproxy/config.json:/etc/toxiproxy/config.json:ro
    deploy:
      resources:
        limits:
          memory: 64M
        reservations:
          memory: 16M

  # OCPP 1.6-J CSMS (optional — docker compose --profile ocpp up)
  # Phase-50 / p50-ocpp. Accepts WebSocket connections from non-Tesla
  # chargers (Wallbox, OpenEVSE, EVlink) on the `ocpp1.6` subprotocol
  # and routes messages through internal/ocpp.Dispatcher.
  #
  # Foundation PR uses the in-memory session store; a Postgres-backed
  # store can be wired in a follow-up without touching the protocol
  # layer.
  #
  # Ports:
  #   9090 - OCPP WebSocket endpoint at /ocpp/<chargePointId> + /healthz
  ocpp-server:
    build:
      context: .
      dockerfile: Dockerfile.ocpp-server
    container_name: teslasync-ocpp-server
    restart: unless-stopped
    profiles:
      - ocpp
    ports:
      - "${OCPP_PORT:-9090}:9090"
    environment:
      - OCPP_LISTEN_ADDR=:9090
      - OCPP_HEARTBEAT_INTERVAL=${OCPP_HEARTBEAT_INTERVAL:-300s}
      - OCPP_READ_DEADLINE=${OCPP_READ_DEADLINE:-900s}
    deploy:
      resources:
        limits:
          memory: 64M
        reservations:
          memory: 16M

volumes:
  postgres_data:
  grafana_data:
  mosquitto_data:
  redis_data:
  prometheus_data:
  pyroscope_data:
  vehicle_photos: